Dynamic Time Series Cycle Analysis And Forecasting Model

Knowledge Discovery in Database is a rather active research domain that involves statistics, database, etc. It is an unusual process that finds out the effective, fresh, latent-effective mode which can be finally apprehensible. The Dynamic time series period analysis and prediction model analyses a serial-typed time series from the point of statistics, finding out the law. thereby succeeding in predicting the future.In this paper the author first introduces a kind of prediction theory梩he multiplayer-transfer method , which was proposed by Zhigang. Han. an domestic expert on cybernetics. It is the result of applying the system-distinguished method out of controlling theory to prediction. Spurning the stationary parameter prediction model, the multiplayer- transfer method regards the predicted variable as a random, dynamic, time-changing system. It decomposes the time-changing status variable prediction into two steps: first, predicting the time-changing parameters; second, on the basis of the first step, predicting the system status variable. The first step will lead to the decrease of the prediction error.But with the multiplayer-transfer prediction model, one of the most important bases in measuring the effect of individual factor to the prediction model is the comparative value among the factors. Conversely, the degree of the factor influencing the prediction variable has been neglected to a large extent. In regression analysis, generally speaking, the more intimate a factor is with the prediction variable, the more effective it will be in the prediction model, as is the very fault of multiplayer-transfer method. The multiplayer-transfer-regression analysis and prediction model combines the regression analysis model to the multiplayer-transfer method.Time series period analysis is an important branch of modern statistics. The multiplayer-transfer method is a relatively ideal statistic method at present, both from the point of theory and that of application. Thus this paper puts forward the dynamic time series period analysis and prediction model. It combines the basic principle of the stepwise regression period analysis to the multiplayer-transfer method. It can not only effectively select every latent period of a time series, but also take advantage of the selected latent periods to make a long-term prediction.